General FAQs

How do I interpret the odds ratios given from JMP?

Beginning in JMP 7, detailed odds ratios for nominal effects are provided to compare any level of the effect to any other level.  These odds ratios are interpreted as the odds of the event increase by a factor of the odds ratio when going from one level to the other level in the nominal effect.  Unit odds ratios are easily interpreted for continuous effects as the odds of the event increase by a factor of the odds ratio for a unit increase in the continuous effect. 

Earlier versions of JMP Software compute odds ratios as: exp((Xmax-Xmin)*beta)) where ‘Xmax’ and ‘Xmin’ are the maximum and minimum values of the "X" effect and ‘beta’ is a parameter estimate for “X”. For continuous effects, the odds of the event increase by a factor of the odds ratio for an increase in the continuous effect going from the lowest value to the highest value. In many cases, a ‘step’ this large is not desired. In this case, one could compute an odds ratio by hand as exp(parmest) which can be interpreted as the increase in the odds of the event given a one unit increase in the effect, or for given values a and b (a>b) of “X”, exp(beta)**(a-b).

If, however, the effect is listed as Nominal, JMP 6 and earlier versions will internally code the levels to 1's and "-1"'s, so, the odds ratio will be computed as OR=exp(2*beta). For nominal effects, the odds ratios given in JMP are not easily interpretable. Since the parameter estimates given in this case are deviations from an overall mean, the odds ratio compares the actual level to a value that is two times the distance from this overall mean. This interpretation isn't too useful. If you create the dummy coding for the effect with additional columns in the data table, the odds ratio will be more useful and meaningful. In this case, the odds ratios tell you the odds of your event of interest occurring as you go from the reference level to the specific effect level.

FAQ # 2079
Last Updated: 2007 Sep 05

Browse JMP FAQs

General FAQs

Statistical FAQs

Support Documents

JMP FAQs

Knowledge Base

System Requirements

Alphabetical List of Features

Known Issues

Contact JMP Support

Report a Problem

Update a Report

support@jmp.com

919.677.8008 (US)

International Support via Worldwide SAS Offices